-
Notifications
You must be signed in to change notification settings - Fork 95
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Add Embedding Quantization to QAT module_swap flow #886
base: main
Are you sure you want to change the base?
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/ao/886
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 1430e0e with merge base a4221df (): This comment was automatically generated by Dr. CI and updates every 15 minutes. |
This pull request was exported from Phabricator. Differential Revision: D62664322 |
This pull request was exported from Phabricator. Differential Revision: D62664322 |
Summary: Pull Request resolved: pytorch#886 Adding the embedding quantizer in the same fashion as the other module swap setup. Differential Revision: D62664322
0ac8d53
to
5259084
Compare
This pull request was exported from Phabricator. Differential Revision: D62664322 |
Summary: Pull Request resolved: pytorch#886 Adding the embedding quantizer in the same fashion as the other module swap setup. Differential Revision: D62664322
5259084
to
a087e50
Compare
@@ -965,6 +965,41 @@ def forward(self, input: torch.Tensor) -> torch.Tensor: | |||
self.precision, | |||
) | |||
|
|||
|
|||
def _replace_embedding_4w( |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I'm wondering if this can be added at the user code side, since we are planning to deprecate the module swap API
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Please don't deprecate the module swap API - it's the easiest to work with and extend.
I'll likely have a headache if I wanted to make things work quickly and effectively with the tensor subclass stuff.
If you guys have a few minutes, we can discuss together how to add this to the tensor subclass stuff as well... but...
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
OK, I think keeping multiple implementations of the same thing might be confusing, we can gather all the requirements and decide on the long term plan I think, I'm asking Andrew to take a stab first
Summary: Pull Request resolved: pytorch#886 Adding the embedding quantizer in the same fashion as the other module swap setup. Differential Revision: D62664322
This pull request was exported from Phabricator. Differential Revision: D62664322 |
a087e50
to
1430e0e
Compare
Summary: Adding the embedding quantizer in the same fashion as the other module swap setup.
Differential Revision: D62664322